Participant Sampling and Description Considerations

In the existing PD literature, comprehensive participant descriptions are often lacking which hampers cross-study comparisons and the ability to elucidate the nature of the relationship between discourse performance and relevant disease-specific variables. However, the careful description of participants on variables relevant to performances on cognitive-linguistic tasks, in general, and to PD, specifically, has become even more

118 important within the recent advances of research examining cognition and language impairments in PD.

Sample size. Sample sizes reported in the existing PD discourse literature have been modest. The mean PD sample size among published studies of spoken monologic discourse is 12.1 with a range of 10 to 21 participants. For controls the mean sample size is 14.9 with a range of 8 to 32 participants. Spoken discourse studies, especially those employing cross-genre, multi-level analyses, require substantial resources placing practical limitations on N sizes. Smaller N sizes, combined with single discourse task sampling practices, create challenges in detecting discourse differences characterized by greater within group variability and smaller effect sizes. Moreover, the dynamic nature of the questions posed in spoken discourse studies requires researchers to analyze samples along a continuum of behaviours creating statistical issues of multiple comparison bias further limiting the detection of smaller, yet potentially important, effects. The current sample size of 38 (n = 19/group) is larger than the mean reported among existing studies of discourse in PD. In the current study, a multivariate approach and an alpha level of .10 were chosen to optimize 1-β given the inherent resource

demands of participant recruitment, data collection, transcription, and analysis. As such, the ability to detect smaller sized effects, possibly obscured in other studies, may partially account for differences in observations between the current study and the existing

literature.

Age and education. With few exceptions (Illes, 1989; McNamara et al., 1992), the age and education profiles of participants in the current study were comparable to other studies of monologic spoken discourse in PD (Ash et al., 2011; Ash et al., 2012a; Murray, 2000; Murray & Lenz, 2001). There were no significant differences in the age or education distributions between the PD and control groups. The existing literature in healthy older adults paints a complicated picture of age effects on spoken discourse performances. In healthy adults, Capilouto, Wright, and Wagovich (2005) reported no effect of age for words per minute or CIUs per minute but did report significant

differences in % CIUs between younger adults (M = 22.4(2.2) years) and older adults (M

119 healthy adults, reported that age effects were significant only for efficiency of

communication content (i.e., using a measure similar to % CIUs) among a coterie of microlinguistic and macrostructural discourse variables (Mackenzie, 2000). Moreover, while participants in the 40 to 59 years group were more efficient in expressing content vs. those in the 75 to 88 years group, no effects of age were observed in the 40 to 59 years group vs. 60 to 74 years group. Other studies have reported similar findings suggesting that age has minimal effect on most microlinguistic and macrostructural variables for highly structured discourse tasks except in the old elderly (i.e., > 75 years of age) (Cooper, 1990; Glosser & Deser, 1992; Mackenzie, 2000; Marini et al., 2005; Obler et al., 1994; Wright et al., 2014). The two discourse tasks selected for the current study (Picture and Retell) were chosen because they minimize such age-related effects on discourse performance differences which are more apparent on discourse tasks such as single picture descriptions and personal narratives that encourage tangential language production (Marini et al., 2005). In the current study, it is unlikely that age differences contributed significantly to observed performance differences between PD vs. controls. Nonetheless, careful consideration was given to equating the groups to minimize age effects.

The range for years of education, while typical for the population from which participants were drawn, was fairly large in the current study. Mackenzie (2000) reported that

education has a significant effect on spoken discourse performance for picture description tasks in healthy adults for a variety of microlinguistic and macrostructural measures (Mackenzie, 2000). Mackenzie’s study (completed in the UK) showed that the effects of education on discourse performance were most pronounced for those individuals not completing a minimal level of secondary school with no post-school

professional/specialized training. This was not the case for any of the participants in the current study. As such, education was not expected to have a substantial effect on the interpretation of the data from the current study in the context of the existing PD discourse literature. However, it is interesting to note that in the current study a significant relationship between discourse performance and education was found, but only for the Retell task a finding that will be discussed in greater detail later in the discussion.

120 Motor and disease severity. Although age and education data are

ubiquitous across PD discourse studies, few studies comprehensively describe their participants relative to PD-specific variables characterizing motor and disease severity (e.g., UPDRS-III scores, disease duration). While the degree of and nature of the relationships between disease variables and task performance differs across specific cognitive and language tasks, in recent years multiple investigators have underscored the importance of collecting and reporting core disease variables in the characterization of participants with PD in studies of cognition (Green et al., 2002). Moreover, it is crucial to consider how variables such as disease duration or motor severity interact with discourse performance. For example, performances on verbal fluency tasks among persons with PD decline with increased disease staging (Araujo et al., 2011; Koerts et al., 2012; Obeso, Casabona, Bringas, Alvarez, & Jahanshahi, 2012; Pereira et al., 2009). However, the relationship between verbal fluency performance and motor impairment severity is ambiguous (Pereira et al., 2009; Signorini & Volpato, 2006). The findings of these studies indicate that a single marker characterizing motor/disease severity in PD for the purposes of elucidating patterns of discourse performance is likely not sufficient.

With the noted exception of Ash et al. (2011) and Ash et al. (2012a), the comprehensive reporting of relevant disease variables is a void in the current literature on discourse performance in PD. While 80% of studies reported some form of PD disease staging (i.e., Hoehn and Yahr or Webster Disability Scale values), only 50% reported disease duration and 40% reported motor severity using the UPDRS-III. Moreover, few authors with the noted exception of the Ash et al. studies (i.e., 20% of studies), reported medication profiles (e.g., levodopa equivalents) for participants. Collectively, the existing literature of discourse impairments in PD and the growing body of neuropsychological literature indicate that in studies of discourse the collection and analysis of these PD-specific disease characterizing variables are of equal importance to other more typically collected variables such as handedness, languages spoken, etc.

121 While, the relationship is less definite than that for global motor impairment and

disease duration, there is emerging evidence that side of symptom presentation (i.e., right vs. left) and type of motor symptoms (i.e., akinetic rigid vs. tremor dominant) may play an important role in the profiling of cognitive disruptions (Burn et al., 2006) and specifically discourse performance in PD (Holtgraves et al., 2010) relative to PD subtypes. In the current study, participants were not classified according to side of symptom presentation or symptom subtype for two reasons: 1) the exact constructs for labeling these subtypes remains debated in the literature and 2) the sample size was not sufficiently powered for subgroup analyses. Regarding the first point, much work is yet to be done to validate a reliable construct for defining subtypes in PD which is complicated by several factors such as: 1) the finding that right vs. left symptom presentation is not always clear in individuals with PD who can present with more impairment in the lower extremities on one side of the body but an opposite pattern of involvement in upper extremity motor impairments and 2) individuals with PD can initially present with tremor dominant symptoms which evolve into a more akinetic rigid form of the disease. Recently, the work of Stebbins et al. (2013) formalized identification of PD subtypes based on UPDRS scores. As these subtype constructs are more clearly defined, future studies may wish to capture this important data for consideration in analyses.

The mean UPDRS-III score for the current study (30.26) differed from the means reported in other studies by a range of 6.36 points (Ash et al., 2012b) to 16.59 points (Jaywant & Pell, 2010). Using a previously reported clinically important difference score (Shulman et al., 2010), the participants in the current study demonstrated more severe motor symptoms on a magnitude that was of moderate to large clinical importance (Large CID ≥ 10.8 points) in relation to other discourse studies reporting UPDRS-III scores. However, it is unclear as to whether the differences observed in the UPDRS-III scores among studies are the result of true differences in participant populations vs. systematic error associated with the measuring tool and/or raters. One of the challenges in

comparing the UPDRS-III data among studies is its reliability. While intra-rater reliability is reported as high with ICCs > .90 (Post, Merkus, de Bie, de Haan, &

122 Speelman, 2005; Siderowf et al., 2002), interrater reliability is less robust (Post et al., 2005). Comparative studies demonstrate a large degree of variation in UPDRS-III scores such that two different raters assessing the same individual over two separate research visits report score differences of up to 16 points (Post et al., 2005). Post, Merkus, de Bie, de Haan, and Speelman (2005) reported that level of training (i.e., movement disorders neurologist, nurse, and neurology resident) has a significant effect on interrater reliability for the UPDRS with movement disorders neurologists consistently assigning lower UPDRS-III scores vs. nurses and residents. Moreover, UPDRS-III scores can vary substantially between ‘on’ and ‘off’ states relative to medicine administration. In the current study, steps were taken to ensure that scores were reliable and optimally interpretable relative to the discourse findings including the use of a single rater,

completion of the Movement Disorders Society-Unified Parkinson Disease Rating Scale (MDS-UPDRS) certification exam by the researcher (AR), assessing participants in relatively similar conditions relative to medication timing, and administering the

UPDRS-III (i.e., motor assessment subtest of the MDS-UPDRS) on the same day as the discourse data were collected.

Similarly the Hoehn and Yahr scores reported in the PD discourse literature also are quite variable. With a mean score of 2.45, a median score of 2.5 and a range of scores from 2 to 4 on the Hoehn and Yahr scale, the PD participants in the current study are generally comparable to the larger body of PD discourse literature. The majority of studies

reported average scores between 2 and 3 out of a possible 5 on the Hoehn and Yahr scale. Ash et al. (2011; 2012a) reported a mean Hoehn and Yahr score of 2.2 (SD = .7) for both studies. In contrast, Murray (2000) reported a mean Hoehn and Yahr score of 3 with scores ranging from 2 to 4. Jaywant and Pell (2010) enrolled participants with the least advanced disease with a mean Hoehn and Yahr score of 1.75. Overall, the Hoehn and Yahr scores reported in the existing literature suggest that typical participants in

discourse studies in PD have bilateral disease of mild to moderate severity with varying levels of balance difficulties but remain physically independent. Hoehn and Yahr scores tend to correlate with scores of motor severity on the UPDRS (Goetz et al., 2004). As such, collectively the data from the UPDRS-III and Hoehn and Yahr scores indicate that the participants in the current study had more severe disease than the participants

123 reported in the Ash et al. (2011; 2012a) and Jaywant and Pell (2010) studies, but

potentially less severe disease than those reported in Murray (2000).

In the PD discourse literature, disease duration is quite variable. The mean years with disease in the current study (9.34) was higher than for the participants in Ash et al. (2011), Ash et al. (2012a), Jaywant and Pell (2010), Murray (2000), and Murray and Lenz (2001); but, was comparable in years to Huber and Darling (2011), Illes et al. (1988), Illes (1989), and McNamara et al. (1992). Capturing data relative to ‘true’ disease duration is challenging in PD given that there is a considerable (and variable) pre- symptomatic phase of the disease when the pathology of PD is advancing but physical and autonomic symptoms remain at a sub-threshold level impeding the assignment of a diagnostic label using current criteria (Braak et al., 2003; Postuma, Gagnon, &

Montplaisir, 2010). Research efforts continue to search for reliable and accurate tools for diagnosing PD in its pre-symptomatic state (Postuma et al., 2010). However, until such time as these tools exist, understanding fully the the relationship between disease

duration and discourse performance will remain challenging. Nonetheless, describing PD participants relative to the duration of their PD symptoms, in the context of other disease- specific measures, is informative.

The value of reporting levodopa equivalent dose (LED) in studies of PD is that it provides a marker of the amount of dopaminergic medication required to manage PD symptoms. While not universal, higher LEDs are associated generally with more severe motor symptoms and longer disease duration. Furthermore, levodopa can have varying effects on cognitive and language tasks (Poletti & Bonuccelli, 2013). Although to date, no study has explored ‘on’ vs. ‘off’ effects in spontaneous language performance in PD. As such, LEDs provide a metric for comparing participants across studies and are

frequently reported in PD studies of cognition. Only three prior studies of discourse in PD reported LEDs (Ash et al., 2011; Ash et al., 2012a; Ash et al., 2012b). The mean LED for the current study was 984.37 mg with wide 95% CIs suggesting participants were variable relative to dopaminergic medications required to manage symptoms. The mean LED for the current study is substantially higher than the dosing reported in other studies of PD discourse. The difference in LEDs is consistent with the higher UPDRS-III

124 and higher Hoehn and Yahr scores reported for participants in the current study and may further indicate that they had more severe motor symptoms vs. those reported in the Ash et al. (2011), Ash et al. (2012a), and Ash et al. (2012b). Participant differences in motor severity, reflecting more advanced pathological states of PD, may aid in the interpretation of discrepancies between the current study and the Ash et al. studies that are expounded in the discussion of the results associated with the specific research questions.

Depression. Depression is common in PD occurring in an estimated 30 to 40% of individuals (Schrag et al., 2007). Depression in PD is linked to impairments in

cognitive processes, specifically executive functions and confrontation naming for nouns (Poletti & Bonuccelli, 2013; Tröster, Stalp, Paolo, Fields, & Koller, 1995). The presence of depression in the control cohort was screened via self-reported history. In the PD cohort, the depression screening protocol included self-report, a clinical chart review, and item 3 from the UPDRS-I. It was not reasonable, within the limitations and purposes of the study, to access the medical records of controls. The UPDRS is a measure validated in PD only and, thus, would not have been appropriate to administer to the control group. In the Schrag et al. (2007) review of depression screening tools in PD Part I of the

UPDRS was determined to be sufficient as a “crude screening tool” for depression in PD (p. 1085). Alpert, Rosen, Welkowitz, and Lieberman reported significant correlations between depression scores and measures of productivity and verbal disruption in the spoken discourse of participants with Parkinson disease dementia (PDD). However, only two studies of spoken discourse in PD without dementia used a formal screening protocol for depression. Murray (2000) and Murray and Lenz (2001) used the Hamilton

Depression Scale. An identified limitation of the current study, discussed at length in Section 5.4, is the lack of a formal screening instrument for depression in the control group and the use of a limited measure in the PD group. However the expected potential impact of depression/anxiety factors on the study results may be minimized given that the majority of PD participants were men and that women with PD have a significantly higher risk of depression/anxiety vs. men with PD (Solla et al., 2012; Song, Gu, An & Chan, 2014). Moreover, as discussed in detail in section 5.4, previous studies have not found a significant effect of depression on discourse measures similar to those reported in the current study (Murray, 2010).

125 Global cognition. Researchers demonstrated that global measures of cognition correlate with performances on more discrete language tasks in PD (Bayles, 1990; Colman et al., 2009; Crescentini et al., 2008; Grossman et al., 2000; Henry & Crawford, 2004; Rodriguez-Ferreiro et al., 2010). Within the existing body of spoken discourse literature in PD, standardized measures of global cognition were reported in 63.6% of studies. Another 27.2% of studies included some version of a self-report task for global cognition. Given the potential influence of global cognition on discourse performance and the elevated risk of dementia in the PD population, controlling for cognition and appropriately screening for PDD is important. Recent studies identified which global measures of cognition are most valid in PD wherein the Mattis Dementia Rating Scale-2

(Mattis, 2001) and the PD-Cognitive Rating Scalewere reported to be the more valid and

reliable measures for identifying PD-MCI and PD dementia (Pagonabarraga et al., 2008). The Mini Mental Status Exam (MMSE), a commonly used screening measure for global cognition, is not recommended for use in PD because of its poor sensitivity for detecting PD-MCI and PDD (Marras et al., 2013). Of the currently published literature on

discourse in PD, only 30% of studies (i.e., Murray 2000; Murray and Lentz 2001; Jaywant and Pell, 2010) have used a PD-specific valid and reliable global cognitive measure as the primary screening tool for determining the presence or absence of dementia.

In the current protocol, none of the participants met the criteria for dementia or mild cognitive impairment using the published MDRS-2 norms for both PD and controls, respectively. For the PD group all participants exceeded the PD-specific MDRS-2 cut- off scores for PDD (Llebaria et al., 2008). Moreover, all participants failed to meet the criteria for PD-MCI using the Level I criteria from the recently published MDS Task Force recommendations (Litvan et al., 2012). Cognitive impairment is present in an estimated 30% of newly diagnosed PD cases (Elgh et al., 2009) and the risk of developing PD-MCI and or PDD increases with disease duration and age (Aarsland, Brønnick, & Fladby, 2011). Moreover, discourse performance (i.e., story grammar and productivity measures) correlates strongly with select measures of executive function (i.e., attention, switching, inhibition) in healthy adults (Cannizzaro & Coelho, 2013). Spoken discourse is a sensitive task for early cognitive changes in dementia and other

126 neurodegenerative diseases (Ash et al., 2006; Duong, Tardif, & Ska, 2003; Orange & Kertesz, 2000; Roberts & Orange, 2013; Roberts-South et al., 2012; Zraick et al., 2011). In the current study, where the objective was to create a discriminative profile of

discourse impairment in PD independent of dementia effects, it was essential that the measures used to identify cognitive impairment be specific to PD and in keeping with current published guidelines. While the mean MDRS-2 scores in the current study are comparable to those reported in Jaywant and Pell (2010), they are higher than those reported in Murray (2000) and Murray and Lenz (2001) by a minimum of seven points for the PD group and two points for the control group. Jurica et al. (2001) suggested that a total score change of two points is a clinically meaningful difference on the MDRS-2 (Jurica et al., 2001). With this in mind, the mean MDRS-2 total score for both PD